Portal de Periódicos da CAPES

Demographic and Educational Success of Lineages in Northern Sweden

2017; Wiley; Volume: 43; Issue: 3 Linguagem: Inglês

10.1111/padr.12091

1728-4457

Martin Kolk, Martin Hällsten,

Social Policy and Reform Studies

Population and Development ReviewVolume 43, Issue 3 p. 491-512 ARTICLEOpen Access Demographic and Educational Success of Lineages in Northern Sweden Martin Kolk, Martin Kolk martin.kolk@sociology.su.se Search for more papers by this authorMartin Hällsten, Martin HällstenSearch for more papers by this author Martin Kolk, Martin Kolk martin.kolk@sociology.su.se Search for more papers by this authorMartin Hällsten, Martin HällstenSearch for more papers by this author First published: 22 September 2017 https://doi.org/10.1111/padr.12091Citations: 3AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Reproductive success and socioeconomic status are inherited across generations, both in contemporary societies and in pre-industrial populations. Previous research has tended to analyze historical settings separately, often with a two-generation rather than multigenerational perspective (Mare 2011). In this study we take the perspective of both social and demographic mobility, examining the success of an entire lineage in terms of the number of descendants and the socioeconomic status of these descendants as indicators of the imprint of previous generations. We apply the prospective design where the origin generation is the unit of analysis and assess both reproductive success, defined as the total number of great grandchildren, and educational attainment, defined as the number and proportion of great grandchildren with tertiary education. We examine how reproductive success and educational attainment are affected by the interplay of variation in the timing and quantum of fertility over time and by period changes such as educational expansion. The present study is unique in that it (1) focuses on how demographic and educational success are reflected both in the total number of descendants and in the average success of these descendants; (2) extends back to the nineteenth century, which crucially covers the period before the fertility transition and thus links historic and modern conditions; (3) provides linkage of data from early industrialization with contemporary data; and (4) highlights the importance of calendar time when studying societies that are in flux. We focus on educational attainment (measured as at least two years of tertiary education) as a measure of socioeconomic success as it is easily comparable over time. Educational attainment reflects a critical dimension of human capital both in the early and late twentieth century that can be comparably measured across time, even though selection into higher education has changed significantly over time. We analyze a predominantly agricultural population in the northern Swedish region of Skellefteå during the mid-nineteenth century (born between 1860 and 1879) and follow their descendants until 2007. Our study covers four generations over 150 years. From 1960 we can link our cohorts with registers of the complete population of Sweden, which allows us to follow the descendants of our initial cohort as they disperse throughout the country. Our study begins in the initial phases of the industrial revolution in Sweden and before its fertility transition and extends to the present, allowing us to study intergenerational processes under periods of rapid social change. The changing economic and social conditions have had significant implications for socioeconomic attainment over time and have reshaped the demographic context. Socioeconomic and demographic transformations in Sweden Industrialization and fertility decline have had enormous effects on modern societies. The industrial revolution was thought to eradicate older economic structures and introduce social fluidity (Kerr et al. 1960), enabling individuals to attain social positions very different from those of their parents. It reshaped market institutions, created new social classes, and provided opportunities for upward mobility, though social immobility was still reproduced in the new industrial societies (Erikson and Goldthorpe 1992). The industrial revolution also led to improvements in living standards not only across cohorts, but also across the life course of many individuals. The fertility decline, fueled by social, economic, cultural, and medical innovations, increased per capita income (Kirk 1996) and prompted a shift from quantity to quality in childbearing (Becker and Barro 1988). This shift toward investing in children is an impetus both for further income growth and for social mobility. The degree to which new fertility behaviors are adopted is crucial in determining the trajectory of families as a result both of socioeconomic differences in the adoption of low-fertility norms (Dribe, Hacker and Scalone 2014; Livi-Bacci 1986) and the transmission of fertility behavior across generations (Anderton et al. 1987). Important institutional developments followed in the context of economic and demographic change. Starting in the nineteenth century, education expanded in Western societies to encompass eventually the whole population. Economic and social change was rapid, with many individuals experiencing fundamental socioeconomic transformation during their lifetimes. This means that the time at which people are born will have a strong influence on their future life chances. By examining both the socio-demographic characteristics of our origin population and the importance of calendar time, we demonstrate how temporal context determines socioeconomic and demographic outcomes across multiple generations. Given a number of rapid social changes, we can expect strong monotonic trends in socioeconomic outcomes. From our prospective and multigenerational perspective, the question is what type of lineage an ancestor leaves behind, and, for many outcomes, any socioeconomic gradient can be dwarfed by time trends. For example, in a recent article in this journal, Barclay and Myrskylä (2016) found that advanced maternal age was associated with lower levels of cognitive and physical outcomes, but the secular positive trend in ability resulting from being born at a later date offset possible physiological effects of being born to an older mother. During industrialization and the attendant structural changes such as educational expansion, a ten-year postponement of childbearing can create substantially better outcomes for the child, and such effects can be multiplied across many generations. One of our aims is to estimate the influence of time and show its contribution vis-à-vis other forces. Previous research To understand the factors that determine socioeconomic outcomes of descendants several generations earlier, one should consider both how groups differ in timing and number of births and how these factors are related to socioeconomic status. It is also necessary to examine how socioeconomic status is reproduced across generations and how relevant factors cited above have changed through time. We provide an overview of some previous research, focusing on how socioeconomic characteristics determine fertility, whether family size is associated with socioeconomic outcomes, the degree to which socioeconomic status is transmitted across generations, and whether fertility behavior is also transmitted across generations. Socioeconomic status and its relationship to fertility The positive gradient between socioeconomic status and fertility characteristic of pre-industrial societies was transformed into a negative pattern after the industrial revolution. Recently there is some evidence of a reemergence of a partial positive gradient in contexts such as contemporary Sweden (e.g. Dribe et al. 2014; Skirbekk 2008). In early modern Europe, parity-specific fertility control was largely absent (Coale and Watkins 1986; Knodel 1988), and thus there was little likelihood of differential marital fertility across social groups. In such a context, entry into marriage was a more important determinant of fertility, as well as the source of social differences in reproduction (Boberg-Fazlic, Sharp and Weisdorf 2011; Clark and Hamilton 2006). Socioeconomic differences in mortality might have played an important role in regulating the number of descendants, but the association between socioeconomic status and mortality was weak in early modern Europe (DeWitte et al. 2016; Edvinsson 2004; Edvinsson and Lindkvist 2011). At the societal level, decreasing infant and child mortality was broadly associated with fertility decline, as expected by classical demographic transition theory (Kirk 1996). With decreasing mortality, the inevitable consequence of not reducing fertility is rapid population growth, such as the postwar global population explosion. While this might be true over a longer time scale, it has been hard to find evidence of such associations between mortality decline and fertility decline at the local level over a shorter time span (e.g. Van de Walle 1986). In our study, we examine how lineage size is dependent on socioeconomic status in the eldest generations and whether this is mediated by fertility in the origin generation. The relationship between family size and socioeconomic outcomes of children Researchers have examined whether the number of siblings determines life chances in both historical and contemporary societies. Numerous studies have found that number of siblings has a negative association with educational outcomes (Downey 1995) and labor market outcomes (Björklund et al. 2004). The resource dilution hypothesis explains this with reference to the quality–quantity tradeoff in children (Becker and Barro 1988). The argument is that with more siblings, each gets less exposure to favorable economic and social resources, which then decreases their chances for later success (Anastasi 1956). Some recent quasi-experimental research, however, finds no such effects (e.g., Angrist, Lavy and Schlosser 2010). These studies describe conditions in rich contemporary societies, where both material and time resources are in generous supply. In pre-industrial times with fewer resources and higher fertility, the relative effects should be more marked, and the research on family size during the demographic transition and industrialization shows stronger effects. Van Bavel et al. (2011) found that children in smaller families were considerably more occupationally upwardly mobile in nineteenth- and early-twentieth-century Antwerp. Klemp and Weisdorf (2011) found a large and significantly negative effect of family size on children's literacy in eighteenth- and nineteenth-century England. It seems plausible that negative effects of large family size were substantial in earlier, less wealthy societies with less government concern with welfare. We address this issue by contrasting average and absolute socioeconomic success of a lineage and show how fertility among earlier generations is related to outcomes of their descendants. Long-term effects of socioeconomic status A central question is the extent to which life chances lie beyond individuals' control and the extent to which individuals are able to shape their own outcomes. Contrary to early sociological expectations, industrialization did not eradicate the persistence of social class over generations (Erikson and Goldthorpe 1992). Many researchers nonetheless saw the persistence to be limited to only one generation, meaning that stratification would be greatly reduced over successive generations (Glass 1954; Hodge 1966). Recent studies focused on the latter half of the twentieth century have rejected this notion and have shown that inequality in labor market outcomes persists over at least three generation across a large number of countries, including the US and Germany (Hertel and Groh-Samberg 2014) and Sweden (Hällsten 2014; Lindahl et al. 2015). Studies of social mobility during nineteenth-century industrialization are scarce, but research conducted in the Netherlands (Knigge, Maas and Leeuwen 2014; Zijdeman 2009) and Sweden (Dribe, Helgertz and Van de Putte 2015) has found that social mobility increased during this period. Other studies have extended historical periods into contemporary times. Lippényi, Maas, and van Leeuwen (2013) found that mobility increased both during industrialization and in modern times. Dribe and Helgertz (2016), studying multiple generations in Sweden, found multigenerational correlations in occupational status but not in earnings. Several Swedish studies have examined multigenerational effects in the twentieth and early-twenty-first century, finding evidence of three- and four-generation effects in school grades, education, and occupation (Hällsten 2014; Lindahl et al. 2015; Modin, Erikson and Vågerö 2013). We show how average and absolute success of future generations depends on the eldest generations' socioeconomic status, an association spanning four generations. Intergenerational continuities in fertility The extent to which reproductive success persists across generations has implications for the relative size of descendant groups (Heyer, Sibert and Austerlitz 2005; Kolk, Cownden and Enquist 2014). In contemporary societies there is clear evidence for intergenerational continuities in fertility (Murphy 2013). These continuities have a multigenerational component where the size of the entire family network matters and is largely independent of socioeconomic status (Kolk 2014), and is likely related to shared preferences across generations (Kolk 2015). In pre-industrial societies the relationship is less clear and has been shown to be weak before the fertility transition (e.g. Reher, Ortega and Sanz-Gimeno 2008). The weak relationship in pre-industrial societies is likely related to the fact that marital fertility was largely uncontrolled. Because of the weak socioeconomic gradient in mortality in early modern Europe, combined with the trivial role of mortality for intergenerational reproduction in contemporary societies where such gradients exist, mortality differences likely are less important than fertility differences for multigenerational associations in reproductive success. In our study, we examine how lineage size and fertility are associated over successive generations. Research integrating demography and socioeconomic status Most of the research described above treats intergenerational continuities across two or more generations as a linear process with one parent and one child (or grandchild) in each generation. However, since most couples have more than one child, it is potentially misleading to ignore the role of fertility in shaping socioeconomic reproduction across generations. This is particularly important when comparing the population frequency of a socioeconomic trait over time. Research that simultaneously examines how different groups have different fertility and also different probabilities of transmitting their own socioeconomic traits to their children has revealed how these two dimensions are related (Lam 1986; Mare 1997; Preston and Campbell 1993). Using pre-industrial data from the Qing dynasty imperial lineage and population registry data for Liaoning for the seventeenth to nineteenth centuries, Mare and Song (2014) find substantial persistence in status over more than ten generations. Their approach, which integrates social mobility with demographic behavior, identifies the number of privileged offspring as the most relevant outcome and examines how that number depends on marriage, fertility, and survival. Since the context is historical China, one of the largest drivers of inequality is polygamy, which is exclusive to advantaged positions and vastly increases the number of privileged offspring in successive generations. Using similar data, Song, Campbell and Lee (2015) examined population growth rates and the risk of lineage extinction for male lineages, modeling the (yearly) population growth rate of lineages. They find that initial status largely affects the number of male direct descendants in a given time period, and that much of this advantage is related to extinction rates, rather than to additive fertility growth of highly successful lineages. Both studies assume that the socioeconomic and demographic context was stable over time, a reasonable assumption for China from the eighteenth to the beginning of the twentieth century, but very different from the context of this study and most of the world over the last century. Two studies apply a prospective approach similar to ours from an evolutionary perspective, examining number of grandchildren of descendants born in 1915–1929 in Sweden after the fertility transition (Goodman and Koupil 2009; Goodman, Koupil, and Lawson 2012). They found a strong linear effect of fertility in the first generation on total number of descendants. Data Our study is based on a combination of national-level administrative register data for the second half of the twentieth century until 2007, together with digitized parish data from northern Sweden between the nineteenth century and 1955. The historical data were collected by the Demographic Database at Umeå University and cover the Skellefteå region of Västerbotten County in northern Sweden (see the map in Figure A11) (Alm Stenflo 1994; Westberg, Engberg and Edvinsson 2016). Skellefteå experienced rapid population growth throughout the nineteenth and early twentieth century. In the early nineteenth century Skellefteå was dominated by landholding farmers, but during the late nineteenth and early twentieth century the area industrialized rapidly (Alm Stenflo 1994). As shown in Figure 1, Skellefteå had higher fertility than the Swedish average up to the first few decades of the twentieth century (Statistics Sweden 1999). The large outmigration to the Americas did largely not affect the Skellefteå region. As we show in Figure A2, by 1960 Skellefteå was in large part socioeconomically representative of all of Sweden in occupational structure. Figure 1Open in figure viewerPowerPoint Total fertility rate in Skellefteå and Sweden, 1860–2010 We have information on the complete population of Sweden after 1960, including birth records linking children to their parents starting in 1932. Our data consist of two separate parts that have been linked together. The first part consists of all individuals in six parishes in Skellefteå. These individuals are linked with modern administrative registers covering the complete population. For 1932 to 1955 we have information on individuals in the Skellefteå region and on individuals born in the rest of Sweden derived from the modern Swedish multigenerational registers and duplicate data for many individuals in our study population. After 1955 our data consist solely of information derived from modern administrative registers. Inclusion in the modern registers is conditional on presence in the registers at any point after 1960. The final year for which we have data is 2007. After 1947 linkages of families are virtually perfect due to unique personal identity numbers. For earlier periods linkages are achieved through names and other identifiers. The population in our historical data is limited to the six parishes. Events (such as births of siblings) observed outside this area are thus not included in our analysis, and this means we to some extent underestimate total numbers of descendants. A consequence of the limited region for our historical data is that we lack information on descendants who migrated outside of Skellefteå. For this reason, we condition our dataset on presence in Skellefteå to age 15 for the first two generations. However, it is still possible that we miss some members of the second and third generation if they migrated outside this region (approximately 10–15 percent of both the second and third generation). For those generations our data are based on a selection of individuals who remained in Skellefteå, and our analyses of descendants in the first generation thus exclude some more mobile descendants (and consequently grandchildren and great grandchildren). We therefore underestimate the total number of descendants of our initial cohort. Variables and measurement Our variable for occupation in the first generation is based on the class scheme from the Demographic Database. We collapsed the original class scheme into five categories that give a meaningful interpretation of the occupational class structure in Skellefteå at the time. The largest group (57 percent) are farmers with ownership of their own land. We group other occupations into agricultural workers (without land ownership) (29 percent) and other non-agricultural workers (6 percent). White-collar workers are a combination of all high-status occupations (just under 1 percent), with many people working as teachers, for the church, or in government positions. In 8 percent of cases we have no information. For our analyses on determinants of the number of descendants and their educational attainment, we use information in administrative registers on educational attainment of great grandchildren (the fourth generation, or G4). These data are retrieved from the 1970 census and from yearly educational registers from 1985 to 2007. We use the highest educational level recorded in any of these sources. Figure 2 illustrates the rising proportions with at least two years of tertiary education in the study area and in Sweden as a whole. We define tertiary education as at least two years of post-secondary education of any kind. To ensure that all individuals have reached an age by which they can be expected to have finished at least two years of post-secondary education, we only include members of G4 who were at least age 27 (born before 1983). These criteria exclude around 7 percent of all G4 individuals. Figure 2Open in figure viewerPowerPoint Proportion with at least two years of tertiary education among great grandchildren of individuals born 1860–1879 in Skellefteå, everyone born in Skellefteå, and everyone born in Sweden, by birth year We analyze the following outcomes and how they are affected by both demographic and socioeconomic factors in the four generations: a) absolute demographic success, measured as the number of descendants in G4 for the 1860–1879 cohorts; b) absolute socioeconomic success, measured as the number of descendants in G4 for the 1860–1879 cohorts having tertiary education; and c) average socioeconomic success, measured as the proportion of descendants in G4 with tertiary education for the 1860–1879 cohorts. We follow only those individuals who were born in Skellefteå and whom we can observe until age 15 in Skellefteå (G1) and who had at least one child (G2) who also stayed in Skellefteå until age 15. The grandchildren (G3) of our first generation are all observed children of G2. They can be part of the historical dataset limited to Skellefteå (before 1955) and/or be part of contemporary Swedish registers (born any time after 1932, conditional on survival to 1960). Our fourth generation (G4) is the great grandchildren of our original cohorts; these are almost always included in the contemporary registers. Our cohorts are chosen to minimize linkage loss through outmigration, while ensuring that we can identify four generations of descendants. Based on the birth years of G3, we estimate that over 99 percent of the completed family size of G3 (observed members of G4) will have been born before 2007. All of our models consist of OLS regressions with G1 as the unit of analysis and outcomes in G4 as our dependent variables. A consequence of using members of G1 as our unit of analysis is that we can only include individual-level characteristics of G1 in our models. Because members of G1 typically had a large number of descendants, we can only analyze factors such as fertility in G3 or birth years in G4 by using group-level statistical measures such as means and quantiles. We can, for example, analyze the 50th percentile of birth years of G4 for a given member of G1. Results Univariate results We begin with descriptive results for our four-generation study population. Figure 3 shows the birth years of the four generations in our study from 1860 to 1983. The range across generations is large and there is substantial generational overlap. For example, the 5th percentile in G4 and the 95th percentile in G2 are very close. In G3 the difference between the 5th and 95th percentile is nearly 70 years. Figure 3Open in figure viewerPowerPoint Box plot of birth years for the four generations in our study, 1860–1983 Figure 4 shows the number of children in each generation. There is a large difference between the fertility of G1 and subsequent generations, as G1 was of reproductive age primarily before the fertility transition, which began comparatively late in Skellefteå. Most members of G2 show post–fertility transition patterns, and both G2 and G3 show fertility levels comparable to the rest of northern Sweden during and just after the fertility transition. TFRs are lower than the TFR cited earlier for the area due to mortality before and during reproductive ages, as well as some undercoverage due to migration. Figure 4Open in figure viewerPowerPoint Box plot of number of children in each of three generations In Figure A3 one can observe intergenerational distances in years for different generational pairs in our sample. The graphs show the median age interval between parents and their children and descendants. The graphs show the large variation in intergenerational age differences that produces the large differences in cohort timing seen in Figure 3. Figure 5 depicts the number of descendants of G1 in G4 (their great grandchildren). Once again, the large variation across members of G1 is striking. The 25th percentile has two great grandchildren, while the 75th percentile has 17. Almost 10 percent of men and women born in 1860–1879 have over 40 great grandchildren. In the other two panels in Figure 5 we show the number and proportion of great grandchildren with tertiary education. Figure 5Open in figure viewerPowerPoint Box plot of total number of great grandchildren of G1 and number and proportion of great grandchildren with tertiary education Figure 6 illustrates the strong positive association between fertility in the origin generation and number of great grandchildren. Large lineages overwhelmingly originate with individuals who themselves have many children. These results contradict the suggestion by evolutionary biologists that low fertility might be associated with future reproductive success among distant descendants (cf. Borgerhoff Mulder 1998). In Figure A4 we examine the degree to which fertility in G1 is associated with fertility levels in G2 and G3. Overall, fertility among descendants appears to be largely independent of fertility in G1, although high fertility in G1 is associated with somewhat higher fertility in G2. Figure 6Open in figure viewerPowerPoint Association between fertility and population size in G1 and number of great grandchildren Multivariate results Because of variations in the timing of childbearing, the long-term descendants of individuals grow up in very different temporal contexts. Given the socioeconomic and demographic changes over the nineteenth and twentieth century, it seems likely that the context in which an individual grows up has a large effect on the chances of both reproductive and socioeconomic success. Using regression analysis, we examine the degree to which factors such as timing and quantum of childbearing across generations, as well as occupation in G1, are independent of each other. Table 1 analyzes our first indicator of reproductive success: the total number of descendants. Since our definition of G1 involves an interval (born between 1860 and 1879) and we have strong time trends, we use the birth year of G1 as a baseline control. Birth timing is central as a larger distance between G1 and G4 is associated with a smaller number of great grandchildren in model 1. Overall, later birth years have a moderate negative effect on eventual number of descendants. The effect is strongest in G1, as these cohorts were in the vanguard of the fertility transition. The effects overall are negative also in later generations as suggested by the slight negative effect of the G1–G4 interval. Table 1. Regression of number of great grandchildren on socio-demographic characteristics of prior generations (1) (2) (3) (4) (5) (6) (7) G1 birth year −0.268*** −0.199*** −0.199*** −0.174*** −0.198*** −0.199*** Birth interval G1–G4 (median) −0.059** −0.099*** −0.105*** −0.100*** −0.101*** G1 no. of children 3.091*** 3.206*** 2.439*** 3.201*** 3.203*** G2 no. of children (median) 4.173*** 4.188*** 4.185*** 4.189*** G3 no. of children (median) 4.804*** 5.278*** 5.273*** 5.251*** G1 Farmer (ref.) 0 0 0 0 G1 White-collar worker −4.285*** −2.106* −1.521* −1.532* G1 Non-agricultural worker −4.132*** −2.187*** −0.524 −0.522 G1 Agricultural worker −0.977** −0.193 −0.228 −0.228 G1 Occupation, other/missing −1.738** −1.002 −0.140 −0.147 G3 Proportion with tertiary educ. (mean) 0.312 No. of individuals (G1) 3812c 4116b 3812c 4765a 3812c 3812c 3812c R-squared 0.0152 0.677 0.674 0.0184 0.263 0.674 0.674 *** p<0.01, ** p<0.05, *p<0.1. a Model specification implies conditioning on at least one member of G2.